程序包 org.apache.flink.ml.common.optimizer

类 SGD

  • 所有已实现的接口:
    Optimizer
    @Internal
public class SGD
extends Object
implements Optimizer
    Stochastic Gradient Descent (SGD) is the mostly wide-used optimizer for optimizing machine learning models. It iteratively makes small adjustments to the machine learning model according to the gradient at each step, to decrease the error of the model.

    See https://en.wikipedia.org/wiki/Stochastic_gradient_descent.

    • 构造器概要

      构造器 
      构造器 说明
      SGD​(int maxIter, double learningRate, int globalBatchSize, double tol, double reg, double elasticNet)  

    • 方法概要

      所有方法 实例方法 具体方法 
      修饰符和类型 方法 说明
      org.apache.flink.streaming.api.datastream.DataStream<org.apache.flink.ml.linalg.DenseVector> optimize​(org.apache.flink.streaming.api.datastream.DataStream<org.apache.flink.ml.linalg.DenseVector> initModelData, org.apache.flink.streaming.api.datastream.DataStream<org.apache.flink.ml.common.feature.LabeledPointWithWeight> trainData, LossFunc lossFunc)
      Optimizes the given loss function using the initial model data and the bounded training data.

    • 构造器详细资料

      • SGD

        public SGD​(int maxIter,
           double learningRate,
           int globalBatchSize,
           double tol,
           double reg,
           double elasticNet)

    • 方法详细资料

      • optimize

        public org.apache.flink.streaming.api.datastream.DataStream<org.apache.flink.ml.linalg.DenseVector> optimize​(org.apache.flink.streaming.api.datastream.DataStream<org.apache.flink.ml.linalg.DenseVector> initModelData,
                                                                                                             org.apache.flink.streaming.api.datastream.DataStream<org.apache.flink.ml.common.feature.LabeledPointWithWeight> trainData,
                                                                                                             LossFunc lossFunc)
        从接口复制的说明: Optimizer
        Optimizes the given loss function using the initial model data and the bounded training data.
        指定者:
        optimize 在接口中 Optimizer
        参数:
        initModelData - The initial model data.
        trainData - The training data.
        lossFunc - The loss function to optimize.
        返回:
        The fitted model data.